DAVID GRAJAL

Cerebra's Qwen3-32B is nuts. Accurate generation and >1000 tokens per second is truly awesome and for the simple queries I like it better than "heavy" models running on standard silicon that take forever to respond. #cerebra #qwen #ai

Using public transportation in Germany is prohibitive for those doing light use. It is around 10 times more expensive than in Spain. Here an example from two tier 2 cities. Is this a death spiral and a sign that a hard reset is needed or just a play of the auto lobby?

I just noticed the last release of #openscad was in 2021 and you are supposed to use the unreleased beta/development versions these days

Somebody reviewed Deepcoder 14B vs Qwen2.5-coder-32B. Qwen2.5-coder 32B seems to be still ahead, over 6 months after release. I have high expectations for Qwen3-coder, we will see.
https://www.reddit.com/r/LocalLLaMA/comments/1jwhp26/deepcoder_14b_vs_qwen25_coder_32b_vs_qwq_32b/

#qwen #ai #codegen

I went to the Nürnberg Data Science & Artificial Intelligence meetup. It was about RAG but I also learned
- How industrial IOT sensors and general IOT infra works
- How the government manipulates unemployment figures with the transformation/training programs

It is always speaking with the people over pizza and drinks where I learn the most #meetup #nuremberg